🍥｜anime

[silk narwhal]

After ... not sure what it is doing making it lower quality

[primal wharf]

Denoising strength?

[silk narwhal]

0.05

[primal wharf]

Try 0.5, idk if that helps

that's what I usually use

[silk narwhal]

lol 0.5 denoise definitly didn't work 😄

finally it is following the inputs

[broken osprey]

Finally, latent couple in comfy ui I can’t show it cus used it for lewd but I’ll get some good ones in q minute

[gusty palm]

https://cdn.discordapp.com/attachments/548263195135967234/1192245869450375258/VID_20240103_181025.mp42.mp4?ex=65a8607e&is=6595eb7e&hm=2497b1c20abf7442af8bc3be5e9217e785f0bf6c13c2949cccd541d7dcccd8e8&

i saw an alien

been a bit since i played with Sdiffusion, anything new?

[worthy delta]

saw a meteor burn up in atmosphere

[gusty palm]

im like 90% sure it was a spacex booster landing cuz thats why i was out there 😛 but im not totally sure

meteor sounds plausible if i didnt have an idea what it is already though

i think i could just barely see the actual rocket launching but it was super cloudy

[silk narwhal]

You wake up at 4am and see this

[worthy delta]

smash

[broken osprey]

Hello, sorry for the ping!! I was wondering if you could tell me how you produced this scene. I was trying to replicate it with latent couple but it isn’t working rhat well. So I wanted to ask you directly!!

Of course if you don’t want to that’s fine too ^^

[primal wharf]

Base prompt + negative base prompt.
The first layer looks like this:

and then add 1 layer per character.

That first layer is the table and chair.

No problem for pinging.

I used regional prompter

not latent couple

https://github.com/hako-mikan/sd-webui-regional-prompter

[broken osprey]

Okay so I’m assuming you can use this too in comfy but I would ahve to look it up: are those masks or just like regions that you paint out for the characters?

[primal wharf]

mask

[broken osprey]

Okay I’ll try it out tomorrow thanks for the info!!

[primal wharf]

Hair crochet

[native halo]

very warm

[stable crane]

I'm not sold by SD Ultimate Upscaler

[broken osprey]

How did you manage such level of control of the characters damn

[stable crane]

oh damn, knitting hair? cool

[nova remnant]

This is...

Pure art

[nova remnant]

how do you come up with that

[stable crane]

I would believe it was hand made

[native halo]

yes a real human bean

[stable crane]

the more I look at it

[brittle rampart]

can i ask? you use comfy?

[stable crane]

@zinc flax king of cuteness

[zinc flax]

Oh no, that's NAI v3

[zinc flax]

Need some cute anime grills here

[brittle rampart]

how you use it? Where?

[stable crane]

you pay $25

novelai

[brittle rampart]

ah... understand...

[stable crane]

I guess there are some cheaper plans too

it's very good, worth it if you have the $

(this message sponsored by #NovelAI)

[brittle rampart]

it like a midjourney

[stable crane]

somewhat, but there's a website and an API

[brittle rampart]

oki-doki

[stable crane]

very anime focused

[zinc flax]

Here's something not NAI v3 lol

[brittle rampart]

it look more cute

for me

[stable crane]

Sleepti has the best cute models

[brittle rampart]

or rather Sleepti using the model from(for) NAI v3? 🤔

[stable crane]

he's been posting mega cute a long time before that came out 😄

[brittle rampart]

argee

[stable crane]

search from: Sleepti has: attachment 🙂

[zinc flax]

I use both 1.5 and v3 actually
Depends on what in the mood I'm at lol

[steady grail]

there is a latent couple on comfyui?

hello guys

[steady grail]

cute

[brittle rampart]

@steady grail can i ask you use this is ?https://civitai.com/models/38784/controlnet-11-models

or this is https://huggingface.co/TencentARC/T2I-Adapter

[steady grail]

let me see

[stable crane]

me on the left

[steady grail]

"STOP! THESE MODELS ARE NOT FOR PROMPTING/IMAGE GENERATION" 🤣

[zinc flax]

D:

[steady grail]

now it's making me curious to see what a control net model can generate lol

maybe just a very confusing noise

i seriously doubt they are tokenized to follow a prompt

i hate t2i adapters. some are very useful, but some generate very weird mutations

[brittle rampart]

The thing is that the usual models for controlnet work very, very slowly for me

[steady grail]

yep, they are very time consuming

i remember that weird mutation where the details where all waves waves.

so damn cursed

i had no way to fix it

[wild marten]

🥰

[native halo]

hand hurts

[steady grail]

try to add (hand muttation:200.000000)... no, maybe all the zeros in the world are not enough to fix this lol

testing controlnet depth on technodrawing i just found out that i'm terrible in prompting with it lol

[steady grail]

cute

[native halo]

yes,(cute hand:3.5)

[wooden coral]

https://www.reddit.com/r/StableDiffusion/comments/18y56y7/a1111_controlnet_updated_with_depth_hand_refiner/
Try this if you get it working

I didnt yet

[steady grail]

why your hands are so cursed? lol

[native halo]

i dont use it cause i want deformities 🤠

[wooden coral]

Sure, you do you

[native halo]

when u lost your hand at a mine explosion in an undisclosed war and after coming back home you wanted a peaceful occupation so you decided to follow your father steps and become an archeologist,while excavating at a location in china you find a suspicious looking desiccated monkey paw that emanates a strange yet luring power so you decide to try your luck and ask it to grant you the wish to grow fingers in your amputated hand so you can recover your mobility and it grants your wish but then you take a look at your hand and see this

[steady grail]

this is the result of every depth controlnet model. the dark one is the pose i used

[nova remnant]

have a look for hand fix using comnfyui https://youtu.be/Tt-Fyn1RA6c

[steady grail]

something is wrong in my way of using this model

[brittle rampart]

but my question was not about the results, but about the operating time. For example, this is the latest generation using controlnet. and this without hires.fix! 7 min!

[brittle rampart]

without controlnet and hires.fix generation time for me 30-40 sec!

and besides, after such a “slowdown”, even turning off the controlnet. the generation speed remains the same for 4-7 minutes. and i have to restart the server

[nova remnant]

whats your vram again? which model version do you use? sd or sdxl? and what resolution do you usually generate images?

[steady grail]

these models are for 1.5

[iron star]

so relatable,happens to me every thursday

[brittle rampart]

i have a 1660 gtx 6 Gb vram and use SD 1.5. SD V1.6.1. contorlnet V1.1.426. my resolutions all time 512x768 ot 544x960.

[brittle rampart]

i think it for 1.5

[nova remnant]

ahh ok, 6gb is bit on the edge for sd. 512x768 is ok. usually with 6gb using controlnet will take up resources

[steady grail]

yup

[brittle rampart]

I also noticed that the first couple of generations proceed at normal speed, in 30-40 sec. but after that there is a sharp “braking” in time.

[steady grail]

the t2iadapter_depth_sd15v2 weighting only 300 mb is obviously the lighter

and from the result, i can see the result is pretty much the same

[brittle rampart]

now i have this is models in folder ControlNet

[steady grail]

there is leak of ram

[brittle rampart]

not more

[steady grail]

nore?

[brittle rampart]

more*

[steady grail]

usually, the first gen is the slower and the rest are faster

[brittle rampart]

the fact of the matter is that my tendency is the opposite

[steady grail]

because something is leaking and freeing memory

[nova remnant]

that's normal when using hi res fix and CN

[steady grail]

something is broken

[brittle rampart]

maybe. but i am not use HiRes.fox step and have a time renegation more 5 min!

[nova remnant]

5 mins w/o hi res fix?

[brittle rampart]

yes!

[nova remnant]

and you are using updated a1111?

[brittle rampart]

now i am on V1.6.1

[nova remnant]

ok here is my suggestion....

refresh the venv folder and run the webui again to update the files

refresh as in delete the venv folder

[brittle rampart]

but this time (5 minutes) is precisely when using controlnet

[nova remnant]

also update the a1111 to the latest version with git pull

[steady grail]

i suppose the leak is in the controlnet extension

or the incompatibility

[brittle rampart]

oki-doki...

1. i am delete Venv folder?
2. start WebUI?

[steady grail]

but torqx solution looks reasonable

[nova remnant]

yes

it will redownload about 2gb file but the point is to clean up any bugs

[brittle rampart]

try...

[steady grail]

yup

it's a pain, but it may be a solution

[nova remnant]

this should be a common practice to refresh venv folder after trying out extensions and other updates

usually in 3 months period

[brittle rampart]

what about this is?

[steady grail]

if it not works we can try scarifying torqx to the ai gods for a solution

[nova remnant]

git pull is console command to update webui

[brittle rampart]

after delete Venv?

[nova remnant]

open your file explorer and navigate to stable-diffusion-webui folder first

you can run the git pull after refershin venv

[brittle rampart]

done

[nova remnant]

run the webui now

once it redownload then run git pull

[brittle rampart]

crap! I ran this before loading WebUI

[nova remnant]

are you getting any error?

redownload first, then run git pull

[brittle rampart]

now i see it

[nova remnant]

ok let it run

[brittle rampart]

yes! loading now

[nova remnant]

webui loading?

[nova remnant]

ok good

it will take few mins

[brittle rampart]

oki-doki.. coffee time!

[nova remnant]

once all that's done, can you show what you have in your webui-user.bat file for command line arg

[brittle rampart]

of course now in bat a have this is

set PYTHON=
set GIT=
set VENV_DIR=

set CUDA_MODULE_LOADING=LAZY
set NUMEXPR_MAX_THREADS=8
rem set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:24

set COMMANDLINE_ARGS=--xformers --no-half --medvram --no-hashing --api-server-stop --medvram-sdxl --opt-sub-quad-attention --opt-split-attention-v1

rem set COMMANDLINE_ARGS=--medvram --xformers --no-half --opt-sdp-attention --upcast-sampling
rem --always-batch-cond-uncond --medvram 


rem set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128

call webui.bat```

[nova remnant]

rename that file with _ before name

one sec.. let me share the content to use and see how that works

try with that

[brittle rampart]

ok. while I'm still downloading something

[nova remnant]

dont mess with the bat fle while its in active session

[brittle rampart]

of course

[nova remnant]

let the files download and when it completes, shut down the webui and close the console, make changes to the bat flle and rerun

[steady grail]

well, a bat is just like an exe.
unless you don't start it, it should be inert

[brittle rampart]

i am just copy it

[steady grail]

i don't think that's a problem

[brittle rampart]

what aboyt this is

[nova remnant]

do that later

[steady grail]

to solve this, you must sacrifice torqx on fire

and for a good harvest too

obviously!

[nova remnant]

btw there is no point in running --medvram-sdxl

remove that from command line

sdxl wont work too good with 6gb

[brittle rampart]

i know about this is. for me SDXL - at the moment from the realm of fantasy

[nova remnant]

yeah thats for sdxl only

but for sdxl you need 8gb at the least

[native halo]

you can run it with 6gb but the 1660 sucks

[nova remnant]

well 6gb is below rec

going to be very slow

[brittle rampart]

yep... Hooray! he's alive! but this is V1.7.0

[nova remnant]

ok good

close those down, and edit the bat file and relaunch

[brittle rampart]

yes

[nova remnant]

you can drag and drop the file i gave

[brittle rampart]

I'm launching

oki-doki

[nova remnant]

ok try generating an image

[brittle rampart]

just last prompt without hires.fix and controlnet?

[nova remnant]

the prompt gen extension you are using has old settings

use hires fix but skip cn for now

[brittle rampart]

this is without Cn and HR

[nova remnant]

ok, try with hires

[brittle rampart]

20 steps

[nova remnant]

also check if you need to update controlnet

[brittle rampart]

try with HR 10 steps

[brittle rampart]

what about this is message?

[nova remnant]

you probably need to update your controlnet

go to the extension tab in your webui then check for update then apply and reload ui

[nova remnant]

shouldnt take this long

[brittle rampart]

with 10 steps HR time 5 min

hmmm

[nova remnant]

that's taking too long

but i remember when i used 4gb nvidia it took me this long too

[native halo]

thats just normal speed for a 1660

[brittle rampart]

before a have 3 min on 20 steps HR

[native halo]

yea the 3min one was that before u updated?

[brittle rampart]

yes

[nova remnant]

update what? a1111?

[brittle rampart]

yes

[nova remnant]

hmm

shouldn't make it slower

[brittle rampart]

3 min with 20 steps HR and with old paramerts in bat file

[nova remnant]

how did you come up with that previous bat file ?

[native halo]

try to use old bat parameters in the new version

[nova remnant]

i see some of the settings are not applicable

steven he was running the previous bat file and getting 5 mins render time

[brittle rampart]

more 5 min render with CN and no HR

[brittle rampart]

just read the startup documentation A1111

[nova remnant]

ok you dont need all that parameters

[native halo]

try to use this params only ```--medvram --xformers --no-half --always-batch-cond-uncond

PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128

[brittle rampart]

oki. try

now my bat ```@echo off

set PYTHON=
set GIT=
set VENV_DIR=

PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128

rem set COMMANDLINE_ARGS=--xformers --medvram --no-half
set COMMANDLINE_ARGS=--medvram --xformers --no-half --always-batch-cond-uncond

call webui.bat```

[brittle rampart]

lost "set"

[nova remnant]

whats this for --always-batch-cond-uncond

i see that's for memory optimization

[native halo]

yea i was reading on github its no longer needed

you can remove it if u want

[nova remnant]

so how long is taking you to render an image w/o hi res or cn?

i think i used to take a break with my 4gb gpu in between each render 👀

but i used hires

[brittle rampart]

well now on last vertions bat and some seed and setting have 5 min render

[native halo]

is that same as before or better?

[brittle rampart]

yes

same as before

i mean before update

on V1.6.1 i have a 3 min

[nova remnant]

some features are deprecated in v1.6.1.

out of use

[native halo]

use v1.6.1 then

[nova remnant]

using old version is not recommended tho, there are features, and updates that will have conflicts

[brittle rampart]

@nova remnant @native halo Well. Let's stick with this version for now. after all, the question was originally about the speed of the controlnet

[native halo]

you cant improve the speed of your 1660 ,3 to 5 mins is normal for that gpu

[nova remnant]

yes but controlnet will not have speedy performance on 6gb

[native halo]

even the 1060 is faster for SD and its an older card

[nova remnant]

one other check, did you try downloading the t2i models? cause regular 1.4gb cn models will eat up memory

[brittle rampart]

now this is my folder Models in CN

[nova remnant]

ok that should help a bit

[brittle rampart]

oh yes i now it. 3-5 min it normal w/o using CN mor me

[nova remnant]

yeah. its soemthing you have to settle with on 6gb vram

[worthy delta]

Any extensions i should get that are handy

[nova remnant]

get adetailer, controlnet, those are kinda essential

[brittle rampart]

btw. how i can use it with CN?

[nova remnant]

the regular way, the t2i models are better optimized

for example to use canny you select canny and load up the respective t2i canny model

[worthy delta]

what is controlnet?

[nova remnant]

https://stable-diffusion-art.com/controlnet/

[native halo]

👆 read that guide it explains everything there

[worthy delta]

an adetailer is a prompt detector?

[nova remnant]

https://learn.thinkdiffusion.com/adetailer-after-detailer-your-automatic-image-enhancement-tool/

[brittle rampart]

can you look on my setting?

[worthy delta]

dayum

[nova remnant]

yeah thats for amazing sweet faces

[worthy delta]

looks pretty good ngl ngl

[brittle rampart]

btw. what about this is https://github.com/ototadana/sd-face-editor

[native halo]

also remember to not use pixel perfect mode,it uses more vram

[brittle rampart]

really?!

[native halo]

yes its slower

[brittle rampart]

i dont know about it

[nova remnant]

havent used it but looks like its quite similar to what adetailer does

[steady grail]

t2i adapters looks more dry

😦

[native halo]

tencent saving the day for SD users again 🙏

[steady grail]

difference in details is evident

[brittle rampart]

soft ages in T2i where?

[native halo]

there isnt

[brittle rampart]

hmmm

whare i can take it?

[native halo]

the same one you were using before

[broken osprey]

What are these models used for??

[native halo]

they also dont have shuffle

[steady grail]

i use sketch or canny

[brittle rampart]

what you mean?

[nova remnant]

for the missing t2i models grab them here https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main

[broken osprey]

Do t 2i models differ from normal control net models

[steady grail]

are pretty similar on the end result

[native halo]

soft edge is better if u want less "edges"

[broken osprey]

I’m not sure what they are used for. Are the models for control net or what

[native halo]

t2i also doesnt have ip2p 😔

[brittle rampart]

againe 1.4 Gb?

[native halo]

downloadn the 700mb ones

[steady grail]

ip2p?

[native halo]

instruct pix2pix

[steady grail]

ohhhh

that's why it loses details

the result is more dry

[nova remnant]

try this https://huggingface.co/webui/ControlNet-modules-safetensors/tree/main

[steady grail]

invalid id main

oh ok

i love depth control net

[brittle rampart]

for softage whiсh file?

[native halo]

its not there

[nova remnant]

you'd have to get 1.4gb for softedge

[brittle rampart]

hmmm

[native halo]

comfy released a bunch of them in smaller size

https://huggingface.co/comfyanonymous/ControlNet-v1-1_fp16_safetensors/tree/main

[steady grail]

cool

i should test them

[brittle rampart]

btw. for using this is t2iadapter_keypose_sd14v1 which Preprocessor can be?

[nova remnant]

use the openpose

im not sure about key pose never used it

[native halo]

only way to make it work in a1111 is to extract keypose from the online keypose preproc gradio demo then add that pic with the extracted pose to controlnet set preporc to none and model to keypose and it will detect the proper keypose

[brittle rampart]

hmm with openpose my geegation time w/o HR againt 5 min!

[broken osprey]

Can someone tell me what t2i stands for? Is it textual inversions or smth?

[brittle rampart]

as I understand it, these are model files for use with controlnet

more "light"

[native halo]

yes lighter versions but prob with less params so the results are not the same

[worthy delta]

what did that?

[nova remnant]

text2image i think

[broken osprey]

Ah si they are no replacement for normal control net models? I have all of the v1.5 ones

[steady grail]

normal deep control net vs t2i

[worthy delta]

which is which?

[nova remnant]

T2I-Adapter is an efficient plug-and-play model that provides extra guidance to pre-trained text-to-image models while freezing the original large text-to-image models. T2I-Adapter aligns internal knowledge in T2I models with external control signals

[brittle rampart]

this is a can check for using this is models?

ot not?

[steady grail]

the first is controlnet 11model, the second t2iadapter 15v2

[nova remnant]

t2i are more efficient and optimized

[nova remnant]

the model lists i shared are meant to work with controlnet method like canny ,depth etc

[brittle rampart]

oki

[steady grail]

this is the lighter model

ill try to upscale, i want to see if it will regain details lost

that's big stuff if it does

this model is way faster

[steady grail]

damn, that's reminds me of the port market

[broken osprey]

But would they affect performance or quality that badly? I mean the big models aren’t that bad

[brittle rampart]

how you use IP-Adapter!!?!? my result ugly

[nova remnant]

What is the difference between T2I-Adapter and ControlNet?
T2I-Adapter vs ControlNets

ControlNets will slow down generation speed by a significant amount while T2I-Adapters have almost zero negative impact on generation speed. In ControlNets the ControlNet model is run once every iteration. For the T2I-Adapter the model runs once in total.

you should be able to google it to see how t2i adapters are better optimized

[nova remnant]

looks like the model you are using cant render the detail too well

[native halo]

maybe too much weight,try to reduce it to 0.7

[brittle rampart]

oki try...

[brittle rampart]

and try use relistic model

[nova remnant]

a brunette girl, flowy hair, in short summer dress, intricate pattern on dress, at a flea market, looking at items on display in the stalls, setting sun, orange hue in the sky.

[silk narwhal]

close enough

[native halo]

nice 🦶

[brittle rampart]

not much better

[native halo]

try with euler a / dpm 2m / ddim sampler and cfg 6 or 5

[brittle rampart]

btw. ADetaler can be activate when im use IP_Adapter?

[native halo]

yes

you need to enable a setting in controlnet

[brittle rampart]

here?

[native halo]

no in settings tab > controlnet > check "allow other scripts to control this extension" box

[brittle rampart]

because the result was not at all what I expected

[brittle rampart]

oh oki!

[brittle rampart]

no better...

[native halo]

face looks good to me

[brittle rampart]

but too young. During generation, Adetailer has not started yet, the face looks very similar to the original

[native halo]

oh try to restart webui and try again

[brittle rampart]

oki

[brittle rampart]

The good news is that the regeneration time is not as slow as it was initially..

[steady grail]

that's a hand controlnet? what?!!!

that's mind blowing

[native halo]

where

[broken osprey]

Quick question, do we need to install this https://github.com/wenquanlu/HandRefiner/tree/main to run the new hand depth control net model

Or should I just download true pruned model and put it alongside the other control net model

[brittle rampart]

what a young Doctor!

[brittle rampart]

now better result... but hand!!!!

[native halo]

thats just normal

either keep generating or inpaint/use depth hand/ hand img / etc

[steady grail]

doctors are all English, except the one that was Scottish.... Anyway, a Doctor can't be Asiatic, it's absurd!

[brittle rampart]

im very love this is BBC serial!

[steady grail]

yeah me too XD

too bad after a certain point, it became too much main stream oriented

[vital raptor]

asia has a north too 😏

[steady grail]

not that an asian doc wouldn't be funny XD

[brittle rampart]

I watched it from the first Doctor to 15

[vital raptor]

can't be any worse than the last 2

[steady grail]

uahahah

[brittle rampart]

15 I couldn’t watch unfortunately

[steady grail]

for first, do you mean the original doctor who from the 60?

[brittle rampart]

so for me it ended with the 14th Doctor

[brittle rampart]

yes

[steady grail]

i found the versions before the bbc stuff too much toony

i couldn't digest the comedy. what i missed?

[vital raptor]

had to look up the 14th doctor and was surprised to see they brought back tennant for a few special episodes.. i was tempted to watch them but saw donna came back too 🤮

worst companion ever, but best doctor

[silk narwhal]

modernizing such an old show will have flaws 😄

[steady grail]

donna was so much funny.

[vital raptor]

i stopped watching after 11, capaldi's doctor was boring

completely changed the character, like he forgot how to be him

[steady grail]

calling her Karen would have been more appropiate lol

[vital raptor]

which is too bad, because clara is super cute

[brittle rampart]

but i love Capaldi`s and Tennat Doctors...

i dont know why....

[vital raptor]

my roommate was watching an episode with capaldi a couple days ago, so i stopped and watching a bit, it was like both the doctor and clara were new and had no idea what was going on or anything about each other as if they hadn't already been doing this for 2+ seasons

it was weird

the writing seemed to be pretty crap for the most part

took me a little to warm up to matt smith after tennant for so long, but he's easily my favorite after tennant now

they were both so much fun

[brittle rampart]

If my memory serves me right, this was exactly the idea and course of the previous season, according to which they were essentially strangers.

although I could be wrong

[vital raptor]

that would make sense if that was it

[worthy delta]

I need some help, image is too vibrant with a vae and too washed out without

[silk narwhal]

wished i knew how to add NNCN upscalers to comfyUI .. but you need to add some exe etc

[worthy delta]

i use kl-f8-anime2.ckpt

[silk narwhal]

use kl-f8-anime2.ckpt for SD1.5 models and https://civitai.com/models/152040/xlvaec?modelVersionId=264491 for SDXL

[worthy delta]

Ill try it out

[silk narwhal]

if using ComfyUI alsway good to add a color correct node before the save image

[worthy delta]

a111 user

[silk narwhal]

believe there is a color correction extension for A1111 too

the model forgot the tardis before this last sampler pass so it added it 😄

[steady grail]

cool

[worthy delta]

yo mate what models do you use?

[steady grail]

soushiki

[worthy delta]

whats that one like?

[steady grail]

it's present on civitai

honestly i really love the bgs it generates

[worthy delta]

bgs?

[steady grail]

backgrounds

[worthy delta]

ahh gotcha gotcha

im looking for a rustic one

[steady grail]

it's very centered on oil painting style

should be ideal for you too

[worthy delta]

ill give it a look

looking for one that does this style

[steady grail]

colors are very sharp

no idea how to prompt such sharpness

[worthy delta]

counterfeit did that

but ive noticed its limit on human type

so imma try model murging it

[worthy delta]

king boo

[silk narwhal]

(best quality, high quality:1.3, cinematic, light particles, depth of field, detailed background), 1girl long_hair breasts large_breasts gloves dress jewelry very_long_hair white_hair earrings pointy_ears tongue_out pink_eyes white_dress colored_skin sharp_teeth crown white_skin

model is awpainting 🙂

[broken osprey]

Aw painting is amazing

[broken osprey]

Holy, now that is awesome, while latent couple produces amazing results on its own too, this new regional prompting thingy is also quite great

[broken osprey]

wait I think Im in the wrong section for this sorry\

[silk narwhal]

no biggie ... lots of random stuff gets talked about in this channel ... 😄

[broken osprey]

hahah guess so yeah, also love your style ^^ looks like a cute little waifu

[silk narwhal]

trying to get transparent skin lora to work with my prompt blending workflow

[broken osprey]

how did you get prompt belding? did you use the regional prompting nodes that impact pack has or did you create your own workflow?

[worthy delta]

bro has cooked us a divine plate

what is that model

[broken osprey]

Now its a little bit more animesque hehe

[broken osprey]

proud of this one, no upscale yet (plan on doing that but the composition was followed thoroughly

makes it feel like dall-3 level of control now

[silk narwhal]

edit of a workflow found on https://comfyworkflows.com/

[broken osprey]

JESUS

uh but do you have the json perhaps?

that link pushes me to the general page :c

[silk narwhal]

drop the image i shared on comfy

[broken osprey]

ah fr fr thought It didn't have metadata hehe mb thanks !!

[worthy delta]

PRAISE THE SUN

[broken osprey]

little winter date how about this one eh? dont rate quality rather composition hehe

okay Imma do 343 iterations 3x upscale see if it doesnt die hehehe

[silk narwhal]

🙂

[broken osprey]

NOOOOO I wanted to do it locally

[silk narwhal]

is local 😄

[broken osprey]

You stole it Jk

What is that ?

[broken osprey]

Quick little question why do you need it to 15k??????

[silk narwhal]

O_O 1.18 MB picture to 208 MB O_O

[broken osprey]

that can’t even be displayed on a monitor LMAO

[silk narwhal]

it did a 4x x 4x upscale 😄

[broken osprey]

What’s the app though may I have its name pwease?

[silk narwhal]

upscayl

[broken osprey]

Does it download the model itself or should I use the ones I have in my pc?

[silk narwhal]

has its own custom models

[broken osprey]

How vram hungry is it?

[silk narwhal]

uses Neural Networked Control Net Models .. does not work yet in A1111 or ComfyUI

[broken osprey]

here is my own upscale,I just want to turn on a detailer to give the face more clarity

[silk narwhal]

blergh Epic Launcher is so slow installing games 😦

[wooden coral]

Breakdomain I think, should have meta data

[broken osprey]

okay? soit should still be pretty vram hungry haha, well Ill give it a try later on today :)

[worthy delta]

Man that one is amazing imma yoink that

[wooden coral]

Could be that you can't find it anymore, some of the breakdomain models got deleted

[broken osprey]

breakdomain could be thbe answer butr I have never seen it produce such vibrant colors, I have it installed and tbf I think it needs an special vae

[silk narwhal]

uses up less than 800 MB of VRAM

[broken osprey]

daym

[steady grail]

that hand controlnet doesnt work for me 😦

[broken osprey]

:c where did you get the model from? I wanna try it too

[wooden coral]

I think I'm using kf, but you can check in the meta data

[steady grail]

https://www.youtube.com/watch?v=Tt-Fyn1RA6c

[silk narwhal]

the meshgraph one .. yea get errors too

[broken osprey]

hmmmm what about using controlnet for the hands, like if you want that maybe idk can you turn only certain body parts in the image?

let me try that for the hand in my image

[silk narwhal]

most onnx related as my system refuses to install onnx dependencies

[wooden coral]

Btw the new depth control net for hands works really well

[steady grail]

all the stuff is active, but simply doesn't hook or generate hands

it's weird as a workflow

[wooden coral]

I advise everyone to try it

[steady grail]

and when it hooks something, have a weird form

[broken osprey]

which one the pruned version from hugging face? or the mesh thingy one

[wooden coral]

I just updated ADetailer and use it in there

Should also pop up with the controlnet update

Not the mesh thing, I think that's unrelated

[broken osprey]

where did you get the model, mind giving me a link to it?

[wooden coral]

It downloads it when using the extension so I don't have a link on me

[silk narwhal]

you can download it in the comfyUI manager: control_sd15_inpaint_depth_hand_fp16.safetensors

note it only works with SD1.5 not yet with SDXL

[broken osprey]

but how about for webui :c

Im trying to come back to the webui, use it for something aside from just storing models lmao

[silk narwhal]

download with manager and move it from the comfy folder to the A1111 folder 😄

not sure if A1111 controlnet is updated for it

[broken osprey]

hahah frfrfr, well I might try it, but how did you get it to work ther,e did you just use a normal loader for ultralitics and chose the new depth hand,

or loaded it through ControlNetLoader node?

[silk narwhal]

i have it controlnet loader > apply controlnet > ydetailer (face) > ksampler > ydetailer (hands) > ksampler > save image

whole workflow arranged left to right .:D

so much pre loaded stuff 😄

[broken osprey]

Wait so like what is that ydetsiler? Which pack does it come in

Also impainting hands is a PAIN

[silk narwhal]

got a link to ydetailer in the notes node

just plop it in custom_nodes and restart ComfyUI

[broken osprey]

jeeez that file is scary, so that installs it alone? hmmm I wonder if it will update through the manager too if I do this

[silk narwhal]

creator of it doesn't have a github ..so doubt it

[broken osprey]

so where did you get itt?

[silk narwhal]

it's pretty much a optimized version of the facedetailer node

Project AI discord

[broken osprey]

Alright Let me finish up my little test of regional prompting and Ill try that one out

[steady grail]

this is weird, i have this node that is in conglict with itself...

[silk narwhal]

what node?

[steady grail]

https://github.com/WASasquatch/PPF_Noise_ComfyUI

Conflicted Nodes:
Blend Latents (PPF Noise) [PowerNoiseSuite], Cross-Hatch Power Fractal (PPF Noise) [PowerNoiseSuite], Images as Latents (PPF Noise) [PowerNoiseSuite], Perlin Power Fractal Latent (PPF Noise) [PowerNoiseSuite]

i feel i should drop this node

[silk narwhal]

Two different suits

[broken osprey]

what do those fractal nodes do?

[silk narwhal]

one is from WAS other is PNS

[steady grail]

it's a better version of normal noise

images are more clear, and you can set a little the color of the image generated

[silk narwhal]

fractals .. everyones favorite math 😄

[steady grail]

ehehehe

[silk narwhal]

as far as i understand regular noise generation randomly asigns a value of 0-255 of a pixel while fractal can do 0.00000000 - 255.00000000

pretty much generating subpixels noise, allowing much smaller details

[vital raptor]

was trying for bigger fox ears since this model makes them look more like cat ears sometimes.. not disappointed 😄

[brittle rampart]

Fennec!

just one face

[sonic portal]

how make??

gib recipe

[silk narwhal]

just had to generate that after someone posted a Joe Tucker gif 😄

[broken osprey]

quick question, in the last part it says clip vision model, which one should I get?

[silk narwhal]

for?

you can click on the blue links in the manager screen and it will take you to the respective githubs with more info

She is large boned

[broken osprey]

Im trying to use these models, but the last one requires that thingy, that clip vision model which Idkl which one it is, I have those models I think already downloaded

[silk narwhal]

it's the Clip-Vit stuff

[silent kraken]

She's actually more of a giant

[silent kraken]

The leg isn't even the scariest part 😂😂

[silk narwhal]

nee-sama waiting for you because you drank her last milk

[stable crane]

reminds me of gremlins

[vital raptor]

I just used 'fennec fox ears' instead of 'fox ears'

[cold talon]

oh hello there

[umbral depot]

Oh hello there

[glossy eagle]

heheehe\

[worthy delta]

Fellas

[worthy delta]

1girl, steampunk, samurai

try it and see yall results

[worthy delta]

@supple raptor What is the most high quality sampler i can use even if it makes my pc lag

[nova remnant]

https://tenor.com/view/kok-cat-gif-25438805

She needs better clothes

I'll bring her to my house so I can give her proper

Clothes

[native halo]

https://tenor.com/view/msyu-anime-cmoney-cute-female-gif-21178494

she looks confortable

[nova remnant]

Wtf it's a real character

[nova remnant]

https://tenor.com/view/scared-cat-running-funny-run-gif-23668419

[supple raptor]

I'd say there's little difference in overall quality between basically all of the DPM++ samplers as well as Restart, and UniPC. Some work better with different prompts and models. DPM++ 2M Karras is the one I usually stick with.

[vital raptor]

ive been using DPM++ 3M SDE Karras with my current sdxl model merges, seems to give the best results, although the 2M samplers are very similar. It really is just "it depends" per model and the look you're going for

I used DDIM exclusively with my custom 1.5 model because it was faster and gave better background details, but it doesn't work with most xl models and also doesn't have the same detail affect

best thing to do is just run a bunch of x/y comparisons on a variety of prompts and find the sampler you like best for the model(s) you're using

[native halo]

small horns

[brittle rampart]

did nee-sama prepare breakfast?

[silk narwhal]

her older sister did

[brittle rampart]

so there are even two of them!?

[silk narwhal]

okay..this hand refiner workflow is wack... it upscales using the upscaler model (which is 4x) then it upscales again by 0.5 before sending in a last ksampler before saving O_O? .. these last steps made the workflow need 18 GB of VRAM

[brittle rampart]

WHY so much?!

[vital raptor]

sounds like it's zooming in on the hand a bunch, generating a cleaner version, then shrinking it back down for the original image

it's easier to make cleaner details when you have more pixels to work with

[native halo]

18gb of vram usage and still looks like someone smudged all over the pic

[silk narwhal]

yea

[brittle rampart]

maybe this is the style?

[native halo]

rubbing oil on paper style

[silk narwhal]

okay .. before the refiner pass

after .. definitely not working correctly 😄

[brittle rampart]

But can someone tell me in general terms the process of adjusting hands using depth maps?

as I see here #🍥｜anime message

[silk narwhal]

you know how controlnet works?

[brittle rampart]

of couse

[brittle rampart]

working with controlnet in my case is a complete pain regarding generation time

[silk narwhal]

can take a bit with loading in the model etc 😄

Prompt executed in 360.04 seconds < for the one above

[brittle rampart]

yeeep..

but.. what about depth map?

[silk narwhal]

depth map .. the further away the pixels are in z-space on the reference the darker it is

so you can get a 3D representation pretty much of a object with depthmaps

the controlnet uses this 3d representation to push the sampler to the same result with enough steps

with realistic images it is easier.. with anime/cartoon harder as there is less 3D shading to work with

[brittle rampart]

I have an idea about the depth maps themselves. but the question is rather how exactly to coordinate the depth map and the generated image for hand correction?

[silk narwhal]

不在

[silk narwhal]

that is where BBoX etc comes in: these are models trained to detect things like people/faces and hands

他们自己是傻瓜，最好离开这个服务器，管理员已经被叫来了 https://cdn.discordapp.com/emojis/962460660397113414.webp?size=48&name=ReimuSchizo&quality=lossless

[lyric zealot]

actually underrated anime

[native halo]

yea too bad its too short,should have been 24 eps 😔

[vital raptor]

which anime is that?

[lyric zealot]

i really really like the announcer sfx

[lyric zealot]

C: The Money of Soul and Possibility Control

[native halo]

Open Deal 🗣️

[lyric zealot]

https://youtu.be/VcmPZ_FX7Ls?t=65

[native halo]

my bank acc after tax evading

[silk narwhal]

anime for the working citizens 🙂 aka taxpayers

hehe when the background layer already looks amazing 😄

[native halo]

just pay your taxes and gamble your money,if u manage to win a lot you will be able to summon a cute girl with your credit card,but if u go bankrupt she will banish 🌚

[silk narwhal]

https://tenor.com/view/msyu-anime-cmoney-cute-female-gif-21178494

O_O so much got altered in the final output from the two initial prompt generations

[steady grail]

doh

morning

yesterday i tested the hand controlnet till late

that stuff needs too much vram and images of very big res to do his job 😦

when it finally hooked on something, the progress was very slow 😦

[silk narwhal]

i removed the double upscale nodes and it worked faster and less vram 😄

[steady grail]

i tested it with a load image as imput

[silk narwhal]

_<

[steady grail]

so, the rresources loaded where already minimal

[silk narwhal]

which is after the hand refiner 😄

[steady grail]

so it just works with very big hands

[silk narwhal]

only good with photorealistic images or generations ..

[steady grail]

yeah, i suppose so

the hand must be clearly visible

it needs a clear hand to hook

[silk narwhal]

it fails to generate a good depthmap for most anime/cartoon/painterly images

[steady grail]

i suppose the best option would be to add a good placeholder to the image, so that it can hook on it

for me is not the best option, considering that very rarely i generate in photoreal

[steady grail]

maybe i found out why it was so slow

i0m getting this error

D:\Git\ComfyUI_windows_portable\python_embeded\lib\site-packages\scipy\sparse_index.py💯 SparseEfficiencyWarning: Changing the sparsity structure of a csr_matrix is expensive. lil_matrix is more efficient.
self._set_intXint(row, col, x.flat[0])
=> loading pretrained model D:\Git\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui_controlnet_aux\ckpts\hr16/ControlNet-HandRefiner-pruned\hrnetv2_w64_imagenet_pretrained.pth
INFO: Created TensorFlow Lite XNNPACK delegate for CPU.

[silk narwhal]

https://openart.ai/workflows/NkhzwEW80FzCcvzzXEsH try this one .. seems to be optimized for 1.5

[steady grail]

i pretty much copied the workflow from this

the problem is with MeshGraphormer-DepthMapPreprocessor

[silk narwhal]

My master workflow is slowly coming together ... full background control, full character control, and now face and hand refinement under control 😄

takes a good 3-7 minutes to gen the final image but got enough earlier image saves 🙂

[steady grail]

wow

another true monster lol

i should add a background controll to

for hand fixin, i suppose i just use adetailer

looks more efficent

for thow, the most funny nodel of the workflow is ipadapter

[silk narwhal]

Prompt executed in 512.15 seconds 😄

[steady grail]

but i need to work a little on a good region prompting

cute

[silk narwhal]

still messed up one hand, cause it didn't detect it 😄

[steady grail]

12%|██████████ | 2/16 [10:44<1:15:40, 324.32s/it]

😂

i'll do it for science lol

[silk narwhal]

O_O .. that is definitly using shared memory

[steady grail]

yep

as the error warned, the model is loaded in cpu

[brittle rampart]

Does anyone have even worse problems with generation time than I do? 😂

[steady grail]

man, i'm working on a 4gvram XD

[silk narwhal]

might have to start with 256x256 😄

[brittle rampart]

I wonder if all this would show 36 hours for me?

[steady grail]

i'm just curious of the end result, i don't expect to implement it on the workflow

[brittle rampart]

I'm certainly a patient person, but not to that extent😂

[steady grail]

the good thing is that i'm not consuming vram, so i can do literally anything else meanwhile the ai is working

[brittle rampart]

how to understand this?

[steady grail]

task manager, and i know it because comfy literally gave me a warning XD

[silk narwhal]

upscaled it for ya 🙂

[brittle rampart]

try

[brittle rampart]

although the size is already quite large

[silk narwhal]

32x .. okay 😛

[silk narwhal]

😄

[brittle rampart]

The main thing! do not lose parts, but purchase additional ones

[silk narwhal]

found the software's max resolution

[brittle rampart]

hmmmm... she made a terrible face

[silk narwhal]

how to tell a upscaler worked well ... if everything is still detailed at 12% zoom 😄

100% zoom at the sleeves

[brittle rampart]

WOW!

[brittle rampart]

there was such a thing, right?!

[brittle rampart]

and what is this mega upscaler?

[silk narwhal]

can you believe a website is charging $100 for 2000 of these kind of upscales while it can be done free local 😄

[brittle rampart]

and what is the secret of the second option?

ComfyUI workflow?

[silk narwhal]

https://www.upscayl.org/

[steady grail]

maybe i should start an upscale site too lol

[brittle rampart]

I understand correctly that this is also based on AI

[native halo]

that doesnt look very good

[silk narwhal]

yes

[brittle rampart]

Why?

[native halo]

if you compare it with other models you would see the difference

the ones from that site looks blurry

and it only has real ersgan,ultrasharp,remacri when there are better and more recent upscalers out there that u can use with either a1111 or comfy

if u want the more recent ones u need to use comfy

[brittle rampart]

comfy on my 6 Gb Vram?

[silk narwhal]

all upscalers currently in Upscayl

[brittle rampart]

is that all that is in this program?

[native halo]

all of those can be used on a111 and comfy

[native halo]

yea i can go up to 32k images but id have to wait like 4hrs idk about you and your driver ver

[silk narwhal]

needs NNCN edited upscale extension or nodes for the same result at the same speed

[native halo]

the best upscaler right now would probably be HAT,i think comfy supports that natively

[steady grail]

hat is good only for realist if i'm not wrong

for anime, i prefer sharp

[native halo]

theres also one for anime 4xLexicaHAT but it was only trained on 43k images from lexica.art images,so it leans towards that style

[brittle rampart]

you all, guy`s know how to confuse

🤪

[steady grail]

lol

[native halo]

swin-fir and swin-ir are second on PNR tho i think those two work on a1111 so u can try them

[brittle rampart]

where i can download it?

[native halo]

swin-ir its included in a1111 when u install it check the upscalers list

[silk narwhal]

Neural Network View 😄

[native halo]

https://tenor.com/view/hacker-pc-meme-matrix-codes-gif-16730883

hackerman

[brittle rampart]

that's psychedelic workflow for me!!!

[steady grail]

who is the mad lad that made this? lol

[steady grail]

wow

[brittle rampart]

scarry

[steady grail]

this resemble so much that anime with an android as protagonist

i'm missing the name, they even made a film to it

[native halo]

alita?

[steady grail]

yeah!

that one!

totally forgot the name

what a masterpiece

i should collect the comic collection

[native halo]

yea the OVA is great the movie is just another washed remake

[steady grail]

yup

that was another anime totally broke me

the stories of the past were so ecceptional

instead, a manga a felt very much was 20 century boys

[native halo]

the 80s and 90s OVAs were always gritty and had a certain charm to it that we dont get anymore 😔

another one that i liked was Lodoss War

[steady grail]

sorry for my broken english, but i got a cold and i'm not very lucid 😦

oh, lodss war, a fantasy master piece

damn!

Warning: Ran out of memory when regular VAE decoding, retrying with tiled VAE decoding.

[brittle rampart]

in tune with the novel Aelita by A. Tolstoy

[steady grail]

now it's restarting with the hand generation

tolstoy! one of my favorite weriters

now that i think about it, i should buy more of his books!

[brittle rampart]

only Alexey and not Lev

2.5x with swin-ir

hands - my pain!

[nova remnant]

Boto

[steady grail]

cute

[wooden coral]

[native halo]

is it normal upscale or high res fix?

[brittle rampart]

img-to-img

[native halo]

try with high res fix at 2x - 0.5 denoising to get more details

[brittle rampart]

oki-doki

[steady grail]

this hand fix control net is ma master piece

only 1 h and 40 min to complete this piece of art

[brittle rampart]

hard pose hand!

[native halo]

speaking of hands

[wooden coral]

Is that with the depth hand refiner?

[steady grail]

i'll continui to fix it by hand

yep

with high resolution and realism is very good

but for it, is very hard to hook too small or too warped hands

or even when the hands are too much connected

[brittle rampart]

but here hands not to bad...

[nova remnant]

[steady grail]

yup

[native halo]

no shes just givin me the middle finger 😔

[steady grail]

i suppose i just try to fix using adetailer and a good prompting

[wooden coral]

[brittle rampart]

spank her!

[brittle rampart]

I’m just afraid that with a denoising of 0.5 I won’t just get added parts, but it might “break” Illustration

[native halo]

it depends on your model sometimes u can up to 0.5 or just stay below 0.4 to keep original img composition

[brittle rampart]

I basically try not to go beyond 0.42

0.36 - 0.38 is normal for me on HiRes.fix step

[native halo]

i always go for 0.5 with normal upscaler and with latent i go with 0.55 or 0.6 then do a batch of 100imgs and delete the ones tha came out deformed 😬

[brittle rampart]

batch of 100imgs!?!!?!?

How long is this?

[native halo]

yea i gen a lot of imgs

like 2hrs

[nova remnant]

[native halo]

my bad its like 4/5 hrs when doing high res fix with adetailer

[nova remnant]

1060 being squeezed dry

[native halo]

75 degrees my room heater

[brittle rampart]

you guys are crazy! Sorryyy 😂

in my case it will be calculated in days

[native halo]

1060 is this big

[brittle rampart]

my reallity

yep...

[wooden coral]

[wooden coral]

Im just running random prompts cause it doesnt really take a lot of time

[vital raptor]

zoom zoom

[brittle rampart]

the geometry is spoiled, the details of the clothes are lost, but the hands are more detailed

[native halo]

what model is that? looks kinda chaotic

[brittle rampart]

https://civitai.com/models/31262/puffy?modelVersionId=98242

[brittle rampart]

https://civitai.com/models/102622 LoRa

[native halo]

oh thats why,its an oil painting style 🤠

[brittle rampart]

yes! i love painting style

[brittle rampart]

@native halo what about 100img? this is first

😂

[brittle rampart]

1660 6Gb Vram

[vital raptor]

😏

100 images? 30 steps each

of course, none of these are any good

[brittle rampart]

who has a gun?

[vital raptor]

lol, these are just 512x512 SD1.5 images using the UniPC sampler and TensorRT to get the fastest possible generation

they're all completely useless

[wooden coral]

[vital raptor]

out of 100 images this is the best one

[brittle rampart]

TensorRT -what is this?

[vital raptor]

https://github.com/NVIDIA/Stable-Diffusion-WebUI-TensorRT

a1111 extension made by nvidia for better tensor performance on 30/40 series cards

[brittle rampart]

Wow! maybe this will save me to some extent?

30|40series....

sad

[vital raptor]

it also takes a lot of HD space, it basically has to make an optimized model, so it takes an extra 2-7GB for each model you generate an engine with it

and those engines don't work with loras, so you have to also generate special engines for loras, and last i checked you can only use 1 lora at a time

[brittle rampart]

i am not have a 30xx card

[brittle rampart]

I'm harsh! my choice is 1660😂

[vital raptor]

My usual gens actually take much longer, around 18-20 seconds.. not currently using TensorRT on SDXL since they take a lot time to generate the engines and I've been switching models a lot testing merges and such

[brittle rampart]

oh I know, I know!!! you need to fix one line in the script set `share=True` in `launch()`. and write me a small sequence of response characters in a personal message!

[vital raptor]

lol

I mean technically I could do that with my 3080TI that's just sitting idle in my server...

[brittle rampart]

this is just a brilliant idea!

Why is there a 30xx series card on the server? is this a render farm?

[vital raptor]

It's what I had before I got the 4090, I put it in my server to run a LLM and TTS for my smart home automation stuff

basically a personal siri/alexa

[vital raptor]

but I haven't finished setting it up so it's just sitting there

[brittle rampart]

I can't live like this!

but you must admit, my idea is brilliant!

[steady grail]

i agree, he is a genius

you should follow his suggestion

[brittle rampart]

he?!? i am a genius!

[steady grail]

yup

[vital raptor]

there are entire businesses built on that exact idea

[steady grail]

ehehehe

i know, right? XD

but it's way less funny

you can't build you custom set up like this

[vital raptor]

tbh, I probably should have just looked into renting GPU time from a service for the times I need it instead of buying a 4090

the 3080TI was just fine for 99% of things :/

[steady grail]

wow

1500$ O.O

[vital raptor]

my 4090 is currently listed for $2475 USD on amazon 😄 I only spent $1750

[steady grail]

as all my set up XD

[brittle rampart]

I’m a good girl and I won’t swear here!

[steady grail]

good girls generate more and talk less XD

[vital raptor]

I use a $1800 tv for my main display 😏 I think I have about ~$8000 in this system total

because I'm single and have nothing better to do with my money

[steady grail]

that's big stuff

don't talk to my about spending big money, i know people that burn much more in gambling -.-'

[vital raptor]

well, gambling is just dumb

[steady grail]

what a lost cause

yup, it is indead

!

[brittle rampart]

2500$!!!

[vital raptor]

I do dumb stuff with my money like buying a tesla 6 months ago because i could, but still not as dumb as gambling 😄

[wooden coral]

mood

I use the odyssey g9 oled

[vital raptor]

nice, I got a 65" LG C1 4k oled tv

[wooden coral]

I feel in love with the ultra wide

[vital raptor]

i was going to get an ultra wide, then i thought "why not tall too?"

[wooden coral]

Its the same with my case, saw it and fell in love

[vital raptor]

replaced my triple 27" setup with the TV, i love it

[wooden coral]

so I had to get it

[vital raptor]

put one of my 27" monitors back up yesterday so i can watch youtube while I'm gaming fullscreen on the TV 😄

[wooden coral]

Thats the only thing I miss right now

a screen to watch videos while gaming

Using my iPad pro for that right now

[vital raptor]

yeah, i'm debating putting up a second one just for discord since i use it a lot for chatting while I game

thought about using an ipad or my laptop or something, but i use headphones most of the time and I didn't want to mess with routing the audio

plus the 4090 has no problems running multiple displays and I still have them around

[wooden coral]

I have a monitor in my hyte Y60 but that aint worth using for discord or anything really

[vital raptor]

meh, I have to keep specifying 'pants' with this model, it likes to show a little too much cheek